Revisiting built environment and travel behavior: A natural experiment accounting for residential self-selection

Numerous studies have attempted to quantify the genuine impact of the built environment factors on travel behavior which is not through residential self-selection (RSS). Nevertheless, the interactions among built environment variables have rarely been included in investigations. To address this gap, this study utilizes a natural experiment formed by the unique urban housing context in Beijing, China. It specifically focuses on resettlement housing families, whose place of residence is largely exogenous, effectively mitigating the bias caused by RSS. By using a gradient boosting decision trees model, this study investigates the combined influence of the built environment on household vehicle kilometers traveled (VKT). The results show that the net built environment contributes about 64% to household VKT, and RSS plays a trivial role (less than 2%) in the relationship between the built environment and household VKT in Beijing. However, the built environment variables influence household VKT in different ways for resettlement households and commercial housing households. These results increase our understanding of how the built environment influences travel behavior.